ABSTRACT
In clinical research, randomization provides the basis for quantitative evaluation of clinical trial data. Randomization is one important technique to avoid bias in a clinical trial. Many clinical trials not only use a permuted block design for randomization, but also do not mask past treatment assignments. This chapter addresses in detail the method developed by Lieven N. Kennes et al. to detect and correct for selection bias in clinical trials. In randomized clinical trials, selection bias is often not regarded as a threat to clinical trial results. It is widely believed that randomization avoids selection bias. For a continuous endpoint, statistical hypotheses tests to detect and correct for selection bias are developed with the method of maximum likelihood. Many clinical trials are not double-blinded. Some situations do not allow for blinding the investigator, while especially late phase clinical trials often abstain from blinding to reflect clinical routine.
